[RASMB] ls-gs, g(s*), and Gaussian fitting

[John Philo]

RASMB,

I've had off-line queries about the simulated data I referred to in
yesterday's post, so I've dug it out. The enclosed graph
(ls-gs_gaussian_fit.jpg) shows what least-squares g(s) in SEDFIT gives for a
simulation (by SEDFIT) of data for 0.52 OD of a single 23 kDa species at
1.696 S at 50000 rpm (this was meant to simulate a real experiment done by
another lab), and a single-Gaussian fit to that distribution. 

As is obvious from the graph, while the main part of the peak does look like
a Gaussian, there is a very distinct shoulder at ~3.4 S. Thus although the
simulation was for a single ideal species, one would definitely (and
falsely) conclude a second species is present.

Although I don't show it in this graph, if you fit this ls-g(s) distribution
to two Gaussians you get 2.2% of a species at 3.37 S (which one would likely
conclude is a compact trimer or extended tetramer).

It is somewhat a matter or semantics whether you say this shows the ls-g(s)
peak shape is non-Gaussian or instead say that ls-g(s) has produced a false
peak. As I recall, in some other testing of ls-g(s) I saw distorted peak
shapes rather than a split peak. In any event, as others have said, the
point is that you really MUST test your analysis methods against known
cases, and simulations of pure species and known mixtures are a
straightforward way to do that.

I've also enclosed a graph showing a one Gaussian fit when the same files
are analyzed to give the g(s*) distribution using the dc/dt approach (file
gs_gaussian_fit.jpg). With that approach you do NOT see a false shoulder at
3.4 S. For sure, though, in detail these g(s*) peaks are not really
Gaussians either, as I published in 2000 [Philo, J. S. (2000). Anal.
Biochem. 279, 151-163]. Further, as that paper shows, when you have
poorly-resolved species a fitting function that is just a "good
approximation" of the peak shape is really not adequate to accurately
quantitate the species fractions, let alone their individual properties.

John Philo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ls-gs_gaussian_fit.JPG
Type: image/jpeg
Size: 72001 bytes
Desc: not available
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20040713/b76e2ea5/attachment.jpeg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: gs_gaussian_fit.JPG
Type: image/jpeg
Size: 54465 bytes
Desc: not available
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20040713/b76e2ea5/attachment-0001.jpeg>